brintos

brintos / linux-shallow public Read only

0
0
Text · 18.5 KiB · b72866e Raw
530 lines · plain
1perf-script(1)2=============3 4NAME5----6perf-script - Read perf.data (created by perf record) and display trace output7 8SYNOPSIS9--------10[verse]11'perf script' [<options>]12'perf script' [<options>] record <script> [<record-options>] <command>13'perf script' [<options>] report <script> [script-args]14'perf script' [<options>] <script> <required-script-args> [<record-options>] <command>15'perf script' [<options>] <top-script> [script-args]16 17DESCRIPTION18-----------19This command reads the input file and displays the trace recorded.20 21There are several variants of perf script:22 23  'perf script' to see a detailed trace of the workload that was24  recorded.25 26  You can also run a set of pre-canned scripts that aggregate and27  summarize the raw trace data in various ways (the list of scripts is28  available via 'perf script -l').  The following variants allow you to29  record and run those scripts:30 31  'perf script record <script> <command>' to record the events required32  for 'perf script report'.  <script> is the name displayed in the33  output of 'perf script --list' i.e. the actual script name minus any34  language extension.  If <command> is not specified, the events are35  recorded using the -a (system-wide) 'perf record' option.36 37  'perf script report <script> [args]' to run and display the results38  of <script>.  <script> is the name displayed in the output of 'perf39  script --list' i.e. the actual script name minus any language40  extension.  The perf.data output from a previous run of 'perf script41  record <script>' is used and should be present for this command to42  succeed.  [args] refers to the (mainly optional) args expected by43  the script.44 45  'perf script <script> <required-script-args> <command>' to both46  record the events required for <script> and to run the <script>47  using 'live-mode' i.e. without writing anything to disk.  <script>48  is the name displayed in the output of 'perf script --list' i.e. the49  actual script name minus any language extension.  If <command> is50  not specified, the events are recorded using the -a (system-wide)51  'perf record' option.  If <script> has any required args, they52  should be specified before <command>.  This mode doesn't allow for53  optional script args to be specified; if optional script args are54  desired, they can be specified using separate 'perf script record'55  and 'perf script report' commands, with the stdout of the record step56  piped to the stdin of the report script, using the '-o -' and '-i -'57  options of the corresponding commands.58 59  'perf script <top-script>' to both record the events required for60  <top-script> and to run the <top-script> using 'live-mode'61  i.e. without writing anything to disk.  <top-script> is the name62  displayed in the output of 'perf script --list' i.e. the actual63  script name minus any language extension; a <top-script> is defined64  as any script name ending with the string 'top'.65 66  [<record-options>] can be passed to the record steps of 'perf script67  record' and 'live-mode' variants; this isn't possible however for68  <top-script> 'live-mode' or 'perf script report' variants.69 70  See the 'SEE ALSO' section for links to language-specific71  information on how to write and run your own trace scripts.72 73OPTIONS74-------75<command>...::76	Any command you can specify in a shell.77 78-D::79--dump-raw-trace=::80        Display verbose dump of the trace data.81 82--dump-unsorted-raw-trace=::83        Same as --dump-raw-trace but not sorted in time order.84 85-L::86--Latency=::87        Show latency attributes (irqs/preemption disabled, etc).88 89-l::90--list=::91        Display a list of available trace scripts.92 93-s ['lang']::94--script=::95        Process trace data with the given script ([lang]:script[.ext]).96	If the string 'lang' is specified in place of a script name, a97        list of supported languages will be displayed instead.98 99-g::100--gen-script=::101        Generate perf-script.[ext] starter script for given language,102        using current perf.data.103 104--dlfilter=<file>::105	Filter sample events using the given shared object file.106	Refer linkperf:perf-dlfilter[1]107 108--dlarg=<arg>::109	Pass 'arg' as an argument to the dlfilter. --dlarg may be repeated110	to add more arguments.111 112--list-dlfilters::113        Display a list of available dlfilters. Use with option -v (must come114        before option --list-dlfilters) to show long descriptions.115 116-a::117        Force system-wide collection.  Scripts run without a <command>118        normally use -a by default, while scripts run with a <command>119        normally don't - this option allows the latter to be run in120        system-wide mode.121 122-i::123--input=::124        Input file name. (default: perf.data unless stdin is a fifo)125 126-d::127--debug-mode::128        Do various checks like samples ordering and lost events.129 130-F::131--fields::132        Comma separated list of fields to print. Options are:133        comm, tid, pid, time, cpu, event, trace, ip, sym, dso, dsoff, addr, symoff,134        srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output,135        brstackinsn, brstackinsnlen, brstackdisasm, brstackoff, callindent, insn, disasm,136        insnlen, synth, phys_addr, metric, misc, srccode, ipc, data_page_size,137        code_page_size, ins_lat, machine_pid, vcpu, cgroup, retire_lat, brcntr,138 139        Field list can be prepended with the type, trace, sw or hw,140        to indicate to which event type the field list applies.141        e.g., -F sw:comm,tid,time,ip,sym  and -F trace:time,cpu,trace142 143		perf script -F <fields>144 145	is equivalent to:146 147		perf script -F trace:<fields> -F sw:<fields> -F hw:<fields>148 149	i.e., the specified fields apply to all event types if the type string150	is not given.151 152	In addition to overriding fields, it is also possible to add or remove153	fields from the defaults. For example154 155		-F -cpu,+insn156 157	removes the cpu field and adds the insn field. Adding/removing fields158	cannot be mixed with normal overriding.159 160	The arguments are processed in the order received. A later usage can161	reset a prior request. e.g.:162 163		-F trace: -F comm,tid,time,ip,sym164 165	The first -F suppresses trace events (field list is ""), but then the166	second invocation sets the fields to comm,tid,time,ip,sym. In this case a167	warning is given to the user:168 169		"Overriding previous field request for all events."170 171	Alternatively, consider the order:172 173		-F comm,tid,time,ip,sym -F trace:174 175	The first -F sets the fields for all events and the second -F176	suppresses trace events. The user is given a warning message about177	the override, and the result of the above is that only S/W and H/W178	events are displayed with the given fields.179 180	It's possible tp add/remove fields only for specific event type:181 182		-Fsw:-cpu,-period183 184	removes cpu and period from software events.185 186	For the 'wildcard' option if a user selected field is invalid for an187	event type, a message is displayed to the user that the option is188	ignored for that type. For example:189 190		$ perf script -F comm,tid,trace191		'trace' not valid for hardware events. Ignoring.192		'trace' not valid for software events. Ignoring.193 194	Alternatively, if the type is given an invalid field is specified it195	is an error. For example:196 197        perf script -v -F sw:comm,tid,trace198        'trace' not valid for software events.199 200	At this point usage is displayed, and perf-script exits.201 202	The flags field is synthesized and may have a value when Instruction203	Trace decoding. The flags are "bcrosyiABExghDt" which stand for branch,204	call, return, conditional, system, asynchronous, interrupt,205	transaction abort, trace begin, trace end, in transaction, VM-Entry,206	VM-Exit, interrupt disabled and interrupt disable toggle respectively.207	Known combinations of flags are printed more nicely e.g.208	"call" for "bc", "return" for "br", "jcc" for "bo", "jmp" for "b",209	"int" for "bci", "iret" for "bri", "syscall" for "bcs", "sysret" for "brs",210	"async" for "by", "hw int" for "bcyi", "tx abrt" for "bA", "tr strt" for "bB",211	"tr end" for "bE", "vmentry" for "bcg", "vmexit" for "bch".212	However the "x", "D" and "t" flags will be displayed separately in those213	cases e.g. "jcc     (xD)" for a condition branch within a transaction214	with interrupts disabled. Note, interrupts becoming disabled is "t",215	whereas interrupts becoming enabled is "Dt".216 217	The callindent field is synthesized and may have a value when218	Instruction Trace decoding. For calls and returns, it will display the219	name of the symbol indented with spaces to reflect the stack depth.220 221	When doing instruction trace decoding, insn, disasm and insnlen give the222	instruction bytes, disassembled instructions (requires libcapstone support)223	and the instruction length of the current instruction respectively.224 225	The synth field is used by synthesized events which may be created when226	Instruction Trace decoding.227 228	The ipc (instructions per cycle) field is synthesized and may have a value when229	Instruction Trace decoding.230 231	The machine_pid and vcpu fields are derived from data resulting from using232	perf inject to insert a perf.data file recorded inside a virtual machine into233	a perf.data file recorded on the host at the same time.234 235	The cgroup fields requires sample having the cgroup id which is saved236	when "--all-cgroups" option is passed to 'perf record'.237 238	Finally, a user may not set fields to none for all event types.239	i.e., -F "" is not allowed.240 241	The brstack output includes branch related information with raw addresses using the242	/v/v/v/v/cycles syntax in the following order:243	FROM: branch source instruction244	TO  : branch target instruction245        M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported246	X/- : X=branch inside a transactional region, -=not in transaction region or not supported247	A/- : A=TSX abort entry, -=not aborted region or not supported248	cycles249 250	The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible.251 252	When brstackinsn is specified the full assembler sequences of branch sequences for each sample253	is printed. This is the full execution path leading to the sample. This is only supported when the254	sample was recorded with perf record -b or -j any.255 256	Use brstackinsnlen to print the brstackinsn lenght. For example, you257	can’t know the next sequential instruction after an unconditional branch unless258	you calculate that based on its length.259 260	brstackdisasm acts like brstackinsn, but will print disassembled instructions if261	perf is built with the capstone library.262 263	The brstackoff field will print an offset into a specific dso/binary.264 265	With the metric option perf script can compute metrics for266	sampling periods, similar to perf stat. This requires267	specifying a group with multiple events defining metrics with the :S option268	for perf record. perf will sample on the first event, and269	print computed metrics for all the events in the group. Please note270	that the metric computed is averaged over the whole sampling271	period (since the last sample), not just for the sample point.272 273	For sample events it's possible to display misc field with -F +misc option,274	following letters are displayed for each bit:275 276	  PERF_RECORD_MISC_KERNEL               K277	  PERF_RECORD_MISC_USER                 U278	  PERF_RECORD_MISC_HYPERVISOR           H279	  PERF_RECORD_MISC_GUEST_KERNEL         G280	  PERF_RECORD_MISC_GUEST_USER           g281	  PERF_RECORD_MISC_MMAP_DATA*           M282	  PERF_RECORD_MISC_COMM_EXEC            E283	  PERF_RECORD_MISC_SWITCH_OUT           S284	  PERF_RECORD_MISC_SWITCH_OUT_PREEMPT   Sp285 286	  $ perf script -F +misc ...287	   sched-messaging  1414 K     28690.636582:       4590 cycles ...288	   sched-messaging  1407 U     28690.636600:     325620 cycles ...289	   sched-messaging  1414 K     28690.636608:      19473 cycles ...290	  misc field ___________/291 292-k::293--vmlinux=<file>::294        vmlinux pathname295 296--kallsyms=<file>::297        kallsyms pathname298 299--symfs=<directory>::300        Look for files with symbols relative to this directory.301 302-G::303--hide-call-graph::304        When printing symbols do not display call chain.305 306--stop-bt::307        Stop display of callgraph at these symbols308 309-C::310--cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can311	be provided as a comma-separated list with no space: 0,1. Ranges of312	CPUs are specified with -: 0-2. Default is to report samples on all313	CPUs.314 315-c::316--comms=::317	Only display events for these comms. CSV that understands318	file://filename entries.319 320--pid=::321	Only show events for given process ID (comma separated list).322 323--tid=::324	Only show events for given thread ID (comma separated list).325 326-I::327--show-info::328	Display extended information about the perf.data file. This adds329	information which may be very large and thus may clutter the display.330	It currently includes: cpu and numa topology of the host system.331	It can only be used with the perf script report mode.332 333--show-kernel-path::334	Try to resolve the path of [kernel.kallsyms]335 336--show-task-events337	Display task related events (e.g. FORK, COMM, EXIT).338 339--show-mmap-events340	Display mmap related events (e.g. MMAP, MMAP2).341 342--show-namespace-events343	Display namespace events i.e. events of type PERF_RECORD_NAMESPACES.344 345--show-switch-events346	Display context switch events i.e. events of type PERF_RECORD_SWITCH or347	PERF_RECORD_SWITCH_CPU_WIDE.348 349--show-lost-events350	Display lost events i.e. events of type PERF_RECORD_LOST.351 352--show-round-events353	Display finished round events i.e. events of type PERF_RECORD_FINISHED_ROUND.354 355--show-bpf-events356	Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.357 358--show-cgroup-events359	Display cgroup events i.e. events of type PERF_RECORD_CGROUP.360 361--show-text-poke-events362	Display text poke events i.e. events of type PERF_RECORD_TEXT_POKE and363	PERF_RECORD_KSYMBOL.364 365--demangle::366	Demangle symbol names to human readable form. It's enabled by default,367	disable with --no-demangle.368 369--demangle-kernel::370	Demangle kernel symbol names to human readable form (for C++ kernels).371 372--addr2line=<path>::373	Path to addr2line binary.374 375--header376	Show perf.data header.377 378--header-only379	Show only perf.data header.380 381--itrace::382	Options for decoding instruction tracing data. The options are:383 384include::itrace.txt[]385 386	To disable decoding entirely, use --no-itrace.387 388--full-source-path::389	Show the full path for source files for srcline output.390 391--max-stack::392        Set the stack depth limit when parsing the callchain, anything393        beyond the specified depth will be ignored. This is a trade-off394        between information loss and faster processing especially for395        workloads that can have a very long callchain stack.396        Note that when using the --itrace option the synthesized callchain size397        will override this value if the synthesized callchain size is bigger.398 399        Default: 127400 401--ns::402	Use 9 decimal places when displaying time (i.e. show the nanoseconds)403 404-f::405--force::406	Don't do ownership validation.407 408--time::409	Only analyze samples within given time window: <start>,<stop>. Times410	have the format seconds.nanoseconds. If start is not given (i.e. time411	string is ',x.y') then analysis starts at the beginning of the file. If412	stop time is not given (i.e. time string is 'x.y,') then analysis goes413	to end of file. Multiple ranges can be separated by spaces, which414	requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"415 416	Also support time percent with multiple time ranges. Time string is417	'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.418 419	For example:420	Select the second 10% time slice:421	perf script --time 10%/2422 423	Select from 0% to 10% time slice:424	perf script --time 0%-10%425 426	Select the first and second 10% time slices:427	perf script --time 10%/1,10%/2428 429	Select from 0% to 10% and 30% to 40% slices:430	perf script --time 0%-10%,30%-40%431 432--max-blocks::433	Set the maximum number of program blocks to print with brstackinsn for434	each sample.435 436--reltime::437	Print time stamps relative to trace start.438 439--deltatime::440	Print time stamps relative to previous event.441 442--per-event-dump::443	Create per event files with a "perf.data.EVENT.dump" name instead of444        printing to stdout, useful, for instance, for generating flamegraphs.445 446--inline::447	If a callgraph address belongs to an inlined function, the inline stack448	will be printed. Each entry has function name and file/line. Enabled by449	default, disable with --no-inline.450 451--insn-trace[=<raw|disasm>]::452	Show instruction stream in bytes (raw) or disassembled (disasm)453	for intel_pt traces. The default is 'raw'. To use xed, combine454	'raw' with --xed to show disassembly done by xed.455 456--xed::457	Run xed disassembler on output. Requires installing the xed disassembler.458 459-S::460--symbols=symbol[,symbol...]::461	Only consider the listed symbols. Symbols are typically a name462	but they may also be hexadecimal address.463 464	The hexadecimal address may be the start address of a symbol or465	any other address to filter the trace records466 467	For example, to select the symbol noploop or the address 0x4007a0:468	perf script --symbols=noploop,0x4007a0469 470	Support filtering trace records by symbol name, start address of471	symbol, any hexadecimal address and address range.472 473	The comparison order is:474 475	1. symbol name comparison476	2. symbol start address comparison.477	3. any hexadecimal address comparison.478	4. address range comparison (see --addr-range).479 480--addr-range::481       Use with -S or --symbols to list traced records within address range.482 483       For example, to list the traced records within the address range484       [0x4007a0, 0x0x4007a9]:485       perf script -S 0x4007a0 --addr-range 10486 487--dsos=::488	Only consider symbols in these DSOs.489 490--call-trace::491	Show call stream for intel_pt traces. The CPUs are interleaved, but492	can be filtered with -C.493 494--call-ret-trace::495	Show call and return stream for intel_pt traces.496 497--graph-function::498	For itrace only show specified functions and their callees for499	itrace. Multiple functions can be separated by comma.500 501--switch-on EVENT_NAME::502	Only consider events after this event is found.503 504--switch-off EVENT_NAME::505	Stop considering events after this event is found.506 507--show-on-off-events::508	Show the --switch-on/off events too.509 510--stitch-lbr::511	Show callgraph with stitched LBRs, which may have more complete512	callgraph. The perf.data file must have been obtained using513	perf record --call-graph lbr.514	Disabled by default. In common cases with call stack overflows,515	it can recreate better call stacks than the default lbr call stack516	output. But this approach is not foolproof. There can be cases517	where it creates incorrect call stacks from incorrect matches.518	The known limitations include exception handing such as519	setjmp/longjmp will have calls/returns not match.520 521:GMEXAMPLECMD: script522:GMEXAMPLESUBCMD:523include::guest-files.txt[]524 525SEE ALSO526--------527linkperf:perf-record[1], linkperf:perf-script-perl[1],528linkperf:perf-script-python[1], linkperf:perf-intel-pt[1],529linkperf:perf-dlfilter[1]530